The research proposed is designed to examine models of human memory and processes of retrieval from memory, and to provide empirical data to test and expand the models. The first models evaluated are global memory models that suppose a test items presented for recognition contacts all of memory to determine an overall value of match by which to discriminate old from new test items. The second class of models is connectionist (neural-like) models and these assume that an item is distributed as a set of features and that there are multiple layers of features (e.g., input and output layers). The proposal describes hypothetical processes of recognition within a connectionist framework and demonstrates problems with the models in accounting for forgetting and learning. Alternative connectionist/neural schemes are proposed as ways of overcoming the problems. Empirical test for both global memory models and the connectionist models are proposed with respect to the time course of availability of different kinds of information (priming, associative, item, and order information). Within the recognition memory paradigm, tests of new view of priming phenomena are also described. Finally, continuing work on comparisons of reaction time retrieval models and extensions from two-choice to multiple-choice models are described. The proposed work is relevant to mental health because of its investigation of the neurally-inspired connectionist models which offer a new framework with which to view various sorts of brain damage such as amnesia resulting from Alzheimer's disease or Korsakoff's syndrome. Other proposed work will advance our understanding of reaction time models and measures, and therefore advance the future possible application of these techniques and measures to understanding brain deficits and diagnosing those deficits.